Secreted factors are proteins

Circadian clocks have been demonstrated to regulate the rhythmic expression of secretory pathway components, as well as the expression of many proteins, including those belonging to tissue-specific secretomes [294], [297]. Additionally, our results suggest that peripheral circadian oscillators communicate via the exchange of secreted signaling molecules. Moreover, by growth factor dependent transcriptional regulation appeared to be a potential mechanism of paracrine communication among peripheral circadian oscillators.

Thus, to test whether active conditioned medium components are likely to be (signaling) proteins, size fractionation in combination with heat treatment, as well as ammonium sulfate precipitation were performed. These experiments aimed to roughly

determine the molecular size and heat sensitivity of active factors, as well as to demonstrate that active factors can be recovered by a standard protein precipitation procedure. Induction of CRE driven luciferase expression was chosen as functional assay rather than phase shifts, or modulation of clock gene expression, due to its rapid, sensitive and reliable read out.

Ultrafiltration using size exclusion centrifugal filters with increasing molecular weight cut-offs (MWCO) showed that active CM factors are retained predominantly in the concentrates of columns with < 50 kD MWCO. For MWCO > 50 kD active factors were found also in the flow through (Figure 3-8 A-B). This may suggest that active CM components are larger than nucleic acid structures, e.g. miRNAs, or small peptides but likely smaller than extracellular vesicles, e.g. exosomes (~100 kD). However, since polyacrylamide gel electrophoresis of CM size fractions showed that filtration column cut-offs are imprecise (Figure 6-4 A,B), the actual size of active medium components should be determined by more sensitive methods. Additionally, regardless of the filter size used, heat treatment (90°C, 10 minutes) abolished CM medium activity of all concentrates and flow throughs tested, suggesting that activity of CM factors is denatured by heat (Figure 3-8 C-D). Thus, based on the molecular size and thermal instability of active medium components, it appeared likely that they belong to the protein class of biomolecules. To further validate that active CM components are proteins, one of the most common protein purification methods was performed. Salting out with ammonium sulfate preserves native protein structures, allowing for the subsequent assessment of precipitate activity. Moreover, while nucleic acids precipitate at low salt concentrations (≤ 30% (NH4)2SO4), proteins are salted out only at higher concentrations [359], enabling the separation of these two molecule classes.

As expected for proteinergic factors, active CM components started to precipitate at (NH4)2SO4 saturations ≥ 30% and largest activity seemed to be recovered from resolubilized precipitates of ~50% ammonium sulfate (Figure 3-8 E-F). Again, these results support the hypothesis that peripheral circadian oscillators communicate by exchange of secreted signaling proteins.

Figure 3-8: Active components in conditioned medium display characteristics of proteins Size fractionation, heat treatment, and ammonium sulfate precipitation of conditioned and control medium was performed to test whether active CM components are proteins. All protein-based methods and stimulation of U-2 OS 7xCRE:Luc reporter cells were performed as described. (A) Raw time series of a representative experiment upon stimulation of 7xCRE:Luc reporter cells with conditioned or control medium size fractions. (B) Quantification of luciferase signal induced by CM relative to control medium (n=2-3 repeat experiment with 3 technical replicates each, mean ± SD, individual values displayed, Unpaired one-tailed student’s t-test against respective concentrate group: *p<0.05, **p<0.01). (C) Raw time series of a representative experiment upon stimulation of 7xCRE:Luc reporter cells with size fractionated and heat treated conditioned and control medium. (D) Quantification of luciferase signal induced by CM relative to control medium (n=2-3 repeat experiment with 3 technical replicates each, mean ± SD, individual values displayed). (E) Raw time series of a representative experiment upon stimulation of 7xCRE:Luc reporter cells with resolubilized ammonium sulfate precipitates. (F) Quantification of luciferase signal induced by CM relative to control medium (n=4-5 repeat experiment with 3 technical replicates each, mean ± SD, individual values displayed).

0

Having gathered indications that active conditioned medium factors are proteins, their identity remained to be determined. However, in a complex mixture like conditioned medium those components modulating circadian dynamics are expected to be present in low abundance. Thus, a two-step chromatography was performed prior to mass spectrometry of conditioned medium in order to enrich active components (again with respect to CRE transcriptional activation). All chromatography and mass spectrometry (MS) experiments were performed in collaboration with our partners at the at the Protein Purification and Analysis Unit of the Max Planck Institute for Infectious Biology (Berlin, GER).

To separate active CM components by size, gel filtration chromatography was performed. In contrast to previous methods, gel filtration chromatography (Superdex HR-200GL column) allows for intermediate-resolution separation of proteins with a fractionation range between 10-600 kD [360]. The protein content of active fractions was determined to identify those with enriched activity compared to CM input, i.e.

strong CRE induction and low protein content. Ultimately, 10 gel filtration chromatography fractions with enriched activity were identified and split into two so-called active CM pools (Figure 3-9 A). Unfortunately, due to the lack of a column size standard, determination of the absolute size of active factors was difficult. However, since active factors eluted early in the chromatography, i.e. they stayed mainly in the mobile phase, their size was estimated to lie within a rather high molecular weight range. To further fractionate and purify active CM components, CM pools were processed by anion exchange chromatography. Ion exchange chromatography fractionates inputs by reversible interactions of charged chromatography columns with charged proteins (based on their pH dependent surface charge). This method is commonly used for high-resolution purification of target proteins [361] but depends on elution of protein factors by a salt gradient (0-1M NaCl). Thus, to avoid salt effects during the 7xCRE:Luc activity assay, all anion exchange chromatography fractions were concentrated and desalted using centrifugal filters with a 3 kD MWCO. Again, the protein content of active fractions was determined to identify those with enriched activity compared to CM input. Both active pools of the initial gel filtration chromatography, showed similar overall activity profiles after anion exchange chromatography (Figure 3-9 B,C). Ultimately, five active fractions per pool with enriched activity were detected (Figure 3-9 B,C).

To identify proteins enriched in these active anion exchange chromatography fractions, as well as in two inactive fractions (as background control) ESI-MS/MS ion trap mass spectrometry was performed by our collaboration partners. Since previous results suggested that peripheral circadian oscillators communicate via paracrine pathways (see above and [61], [293]), resulting protein hits were filtered for human secreted proteins [362]. Those proteins contained in inactive background controls were considered impurities and removed from the list of active proteins, resulting in a limited number of secreted protein candidates (Figure 3-9 D-E). To further reduce the number of resulting hits, proteins were only considered potential “coupling factors” if they were present in at least 7 out of the 10 active fractions. Isoform specificity was disregarded because conventional mass spectrometry approaches often fail to clearly assign protein isoforms based on peptide information [363]. PSG (pregnancy-specific glycoprotein), SFRP (selected frizzled-related protein), SMOC (SPARC-related modular calcium binding protein), and TGFB (transforming growth factor beta) were identified as candidate coupling factors (Figure 3-9 D,E). Interestingly, except for PSGs, all candidate factors have been indicated in direct or ECM dependent cell-cell communication. However, only TGF-b signaling pathways has been described to interact with the circadian clock machinery [335], [339], [340].

Overall, findings indicate that active conditioned medium components are proteins.

Moreover, consistent with previous results, indicating that secreted molecules mediate intercellular coupling, as well as due to its a priori role as paracrine signaling factor and known interaction with the circadian clock machinery, TGF-b appeared to be a likely candidate coupling factor.

Figure 3-9: Active and secreted CM factors identified by chromatography and mass spectrometry To identify active conditioned medium components, chromatography and mass spectrometry of CM was performed by our collaboration partners at the Max Planck Institute for Infectious Biology (Berlin, GER) as described. Fraction activity was determined by stimulation of U-2 OS 7xCRE:Luc cells and resulting protein hits filtered for components of the human secretome (as described in methods). (A) Quantification of recovered activity of gel filtration chromatography fractions relative to input, as well as of protein content (green=active fractions used for anion exchange chromatography, black=inactive fractions, red=protein content determined by BCA assay). (B-C) Quantification of recovered activity of anion exchange chromatography fractions relative to input, as well as of protein content (green=active fractions analyzed by MS, grey=inactive fractions analyzed by MS, black=inactive fractions, red=protein content approximated by absorption at 280 nm). (D-E) Active secreted protein hits identified in anion exchange chromatography fractions #28-32 of gel filtration chromatography pool1 (D) and pool2 (E) (red=hits common to at least 7 active fractions across pool1 and pool2).

15 20 25 30 35 40 45